Complete Tracklength and Collision Estimator Tallies

[bhermanmit]

I would eventually like to do some comparisons between different tally estimators for various applications. Right now I am forced to use full ANALOG tallies or a combination of TRACKLENGTH and ANALOG tallies. We need there to be 3 options: ANALOG, COLLISION and TRACKLENGTH. Therefore no matter what the transport code sampled for a reaction type, we have the ability to re-sample physics when we need to get energy-out and neutron multiplicity information.

Currently, I don't see any elegant way of re-sampling these physics without overwriting global information that would change the transport physics. If we could some how in sample_reaction we have it call other subroutines for different physics. For example, like a subroutine to sample_scatter_reaction, that doesn't overwrite global variables, but returns outgoing energy, multiplicity, angle, etc. Something similar needs to be done for sample_fission_reaction.

Then we could sample every tally reaction type, if that score exists.

[paulromano]

I generally agree this needs to happen at some point. I also want to do some major restructuring in the tally module since I feel there is too much duplication (e.g., just look how many times the select case(score_bin) block appears). This change could naturally go along with that. What time frame do you need/want this on?

[bhermanmit]

It would be nice to get this done soon, just because I am using ANALOG a lot and am not sure if my conclusions that I am making are because of these tallies. If its beneficial for your work as well, I would be happy to brainstorm and help get this coded (and tested with neutron balance!).

[nelsonag]

We may consider this one closed per #455: although that PR didnt add collision estimators for all types, I think an issue specifically for dealing with the remaining scores would be more useful

[paulromano]

@nelsonag Reading Bryan's original description, I believe the intent of this issue was addressing the fact that there are scores/filters that can only be used with analog tallies. Side note: I think we're in a much better state these days with respect to not touching global variables when sampling physics (as well as potential cyclic dependencies), so it probably wouldn't be terribly difficult at this point to implement.